Cryptanalysis

When two people exchange a secret message, they first turn the text into unreadable gibberish. This transformation is called encryption, and normally only someone who knows the matching secret word can reverse it. Cryptanalysis is the attempt to get the plaintext anyway — without that secret word. It is thus the counterpart to encryption: one side builds locks, the other checks whether they can be picked. Importantly, this is not automatically something criminal. Most cryptanalysts work at universities or security firms and publish their findings so that weak methods are replaced in time.

Why encryption only becomes trustworthy through attacks

You cannot prove that an encryption method is secure. You can only establish that no one has found a practicable attack for decades. This is exactly what the trust in the methods that protect your banking transactions and messages today is based on. Without cryptanalysis, no one would know which of them hold up and which merely look complicated.

That is why a strict principle applies in the field: a method must be open to scrutiny. Anyone who keeps their method secret and claims it is secure precisely because of that makes themselves suspect. Experts dismissively call this "security by obscurity," meaning security through concealment. Reliable, by contrast, is a method whose blueprint everyone knows and that still no one breaks.

Historically, cryptanalysis has helped decide wars. In the Second World War, Polish and British mathematicians broke the German cipher machine Enigma. The Germans considered their machine unbreakable and had made a fundamental error in doing so: they underestimated how much structure their own radio messages revealed.

What codebreakers work with

The simplest attack is called brute force: trying out all possible keys. With modern methods, there are so many of these that all the computers in the world would need billions of years. Cryptanalysis therefore looks for shortcuts. Any shortcut that is faster than trying every possibility already counts as a break in research terms — even if it is still too slow to be practical.

A classic tool is frequency analysis. In German, E is the most common letter. Anyone who, with a simple method, counts which character appears most often in the ciphertext has a good first guess. Modern methods defend against this by encrypting identical letters differently depending on their position.

Often, though, experts don't attack the mathematics but the implementation. A device takes minimally longer to compute with the real key, or measurably uses more power. Such side effects can be used to reconstruct the key. These attacks are called side-channel attacks and are, in practice, more successful than purely computational attacks.

Cryptanalysis in today's communications and technology

You benefit from this every time the padlock symbol appears in your browser. The methods behind it were previously attacked publicly and survived. Methods that did not manage this were switched off: the checksum MD5 and the old protocol SSL are now considered unusable because cryptanalysts cracked them.

Currently, much revolves around quantum computers. Such machines could very quickly break certain methods that protect banking data today. In response, researchers are developing so-called post-quantum-secure methods and immediately subjecting them to cryptanalysis. Several candidates have already failed in the process — which is exactly the point of the exercise.

Artificial intelligence is also increasingly involved. Learning programs can find patterns in measurement data that humans overlook, thereby improving side-channel attacks. This does not crack the major mathematical methods, though. A common misconception is that AI would make encryption generally worthless — so far it merely shifts where the weaknesses lie.